Poster Title
Marion Bauman1, Yuhan Cui1, Varun Patel1, Aaron Schwall1
1 Department of Data Science and Analytics, Georgetown University
1 Introduction
Americans are constantly taking trips by commuting, going to the store, visiting friends, or going on vacation. Understanding American travel patterns and why Americans travel can provide insights into fields ranging from environmental protection resource allocation to urban planning to economic trends. This statistical research is aimed at studying travel behavior of the American population by predicting the purpose of a trip.
2 Data
In this study, we used data from the 2022 National Household Travel Survey (NHTS) collected by the U.S. Department of Transportation Federal Highway Administration. This survey data details the trips that Americans take contains information on individual and their travel habits, including demographic information, modes of transportation, and the purpose for travel. It is one of the only data sources that allows users to analyze American travel behaviors based on household and person level information.
3 Modeling
In order to determine the best model for predicting trip purposes, we trained a variety of statistical learning models on the data. First we will discuss each model trained and then we will go over the results and best model.
Logistic Regression
Random Forest
Support Vector Machine (SVM)
Our best SVM model used a learning rate C of 10 and a polynomial kernel. It has an accuracy of ~53.3%.
XGBoost
Our best xgboost model had parameters subsample = 0.6, number of estimators = 300, max_depth = 9, learning_rate = 0.1, gamma = 0, and colsample_bytre = 1.0. It achieved an accuracy rate of ~63%.
Nueral Net
4 Results
Visualize results
5 Conclusions
Draw conclusions
References
Federal Highway Administration. (2022). 2022 NextGen NHTS National Passenger OD Data, U.S. Department of Transportation, Washington, DC. Available online: https://nhts.ornl.gov/od/.